Chicken Scheme 6.0 released

The dominant theme of 6.0 is a hard pivot toward R7RS-small. Primitives that historically lived in (chicken base) and (chicken io) have been moved into (scheme base), the legacy r4rs/r5rs module names are renamed, and the SRFI aliases that those standards used to expose (srfi-0, srfi-6, srfi-9, srfi-11, srfi-23, srfi-39, srfi-98) are deleted because R7RS now covers them. For an engineer, the practical consequence is a wide import-and-rename pass: module names shift, and any code that imported via the old aliases or relied on non-standard locations will not load against the new core libraries.

The UTF-8 string representation is the deepest change. Strings become fully Unicode-capable, but locatives on strings now index by code point rather than byte, and the file and random APIs that used to take string arguments for byte data (file-read, file-write, set-pseudo-random-seed!, random-bytes) now require bytevectors. The blob read syntax (#${...}) is removed outright, so code that parsed with it will fail at read time rather than runtime. The bytevector story does clean up in parallel: (chicken blob) is gone, replaced by (chicken bytevector), and bytevectors are now fully equivalent to SRFI-4 u8vectors, with read-bytevector and write-bytevector added to (chicken io).

System-interface changes are narrower but consequential. The process API (process-fork, process-run, process, process*) now returns a process object instead of a raw PID, so callers must use the new record accessors to get exit status and the input/output ports; process-wait and process-signal accept either. File locking in (chicken file posix) is rebuilt on flock(2), operates over whole files, and is thread-safe, with file-test-lock removed. Path handling is also normalized: library procedures that return paths and make-pathname now always use forward slashes, which is a deliberate choice on Windows that may surprise code that does string-level path manipulation on platform-specific separators.

Smaller breaking changes are scattered through the release. The #ci and #cs read syntaxes are removed in favor of #![no-]fold-case markers. Hexadecimal escape sequences in strings and extended symbols must now be terminated by a semicolon to follow R7RS. define-record-type is now generative per R7RS, which is a semantic shift for any code that depended on record-type identity across compilation units. number->string now accepts bases up to 36 instead of 16. These are individually small, but they add up: a migration script that just renames modules will still miss them, and they are not the kind of change a compiler will flag for you.

Only one feed (Lobsters) carried this event, and the headline framing there was minimal, the summary field reads 'Comments', so there is no editorial angle to corroborate against or contrast with. The note above is built from the release notes extract alone, which limits how much framing can be reported. What can be said is that the adoption cost is meaningful: anything beyond a toy program will need a real migration pass, and the upside is a noticeably smaller Chicken-specific API surface in exchange.